The error looks like this
Exception in thread "Thread-1" java.lang.NullPointerException
at java.util.LinkedHashMap$Entry.remove(LinkedHashMap.java:332)
at java.util.LinkedHashMap$Entry.recordAccess(LinkedHashMap.java:356)
at java.util.LinkedHashMap.get(LinkedHashMap.java:304)
at Server.getLastFinishedCommands(Server.java:9086)
at Server.processPacket(Server.java:484)
at PacketWorker.run(PacketWorker.java:34)
at java.lang.Thread.run(Thread.java:744)
Inside getLastFinishedCommands
I use
public List<CCommand> getLastFinishedCommands(UserProfile player) {
List<CCommand> returnList = new ArrayList<CCommand>();
if(!finishedCommands.containsKey(player.myWebsitecmd-1)) {
getSavedState(player);
return null;
}
try { //<-- added this try/catch so it doesn't happen again.
//Get commands.
CCommand cmd;
long i;
long startIndex = player.myWebsitecmd;
long endIndex = startIndex+LIMIT_COMMANDS;
for(i = startIndex; i <= endIndex; i++) {
cmd = finishedCommands.get(i); //<-- this is line 9086
if(cmd == null) {
return returnList;
}
returnList.add(cmd);
}
} catch(Exception e) {} //<-- added this try/catch so it doesn't happen again.
return returnList;
}
I wanted to make a Map that auto removes old entries so I used this snippet
public static <K, V> Map<K, V> createLRUMap(final int maxEntries) {
return new LinkedHashMap<K, V>(maxEntries*3/2, 0.7f, true) {
@Override
protected boolean removeEldestEntry(Map.Entry<K, V> eldest) {
return size() > maxEntries;
}
};
}
Used it like this
public static int final MAX_COMMANDS_QUEUE = 5000;
public Map<Long, CCommand> finishedCommands = createLRUMap(MAX_COMMANDS_QUEUE);
Obviously it's some kind of CocurrentModifcationException which happens when using with multiple threads.. but why does it crash internally, anyone know how I can use this with like a CocurrentHashMap? I'm trying to fix this without resorting to just putting a try/catch around the whole getLastFinishedCommands
function.
I want a Map that clears itself from old junk but still holds atleast 5000 key/value entries.
You said that multiple threads are accessing this map. This could indeed cause the NPE in the remove
operation of a LinkedHashMap.Entry
instance. This is the implementation of this method:
private void remove() {
before.after = after;
after.before = before;
}
Here before
and after refer
to the linked predecessor and successor of the current entry. If another thread already changed the linking between the entries, this could of course result in an unexpected behavior, such as the NPE.
The solution is - you guessed correctly - to wrap your produced map in a synchronized map. Such as:
public static <K, V> Map<K, V> createLRUMap(final int maxEntries) {
Map<K,V> result = new LinkedHashMap<K, V>(maxEntries*3/2, 0.7f, true) {
@Override
protected boolean removeEldestEntry(Map.Entry<K, V> eldest) {
return size() > maxEntries;
}
};
return Collections.synchronizedMap(result);
}
This synchronized wrapper will indeed synchronize all calls to the underlying map, so only one single thread is allowed to go through each method (such as get, put, contains, size, and so on).